mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-03-07 18:27:40 +08:00
Code Issues Packages Projects Releases Wiki Activity

clean the commented asm instructions because now I'm sure

Browse Source 
the previous fix is ok
This commit is contained in:
Gael Guennebaud 2009-09-17 23:34:00 +02:00
parent f2737148b0
commit 5ba7fe3bee
1 changed files with 1 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
Eigen/src/Core/arch/SSE/PacketMath.h
View File

@ -76,17 +76,14 @@ template<> struct ei_unpacket_traits<Packet4i> { typedef int type; enum {size
#ifdef __GNUC__ #ifdef __GNUC__
// Sometimes GCC implements _mm_set1_p* using multiple moves, // Sometimes GCC implements _mm_set1_p* using multiple moves,
// that is inefficient :( // that is inefficient :( (e.g., see ei_gemm_pack_rhs)
// TODO make sure the new solution using the shuffle/unpacklo is ok
template<> EIGEN_STRONG_INLINE Packet4f ei_pset1<float>(const float& from) { template<> EIGEN_STRONG_INLINE Packet4f ei_pset1<float>(const float& from) {
Packet4f res = _mm_set_ss(from); Packet4f res = _mm_set_ss(from);
return _mm_shuffle_ps(res,res,0); return _mm_shuffle_ps(res,res,0);
//asm("shufps $0, %[x], %[x]" : [x] "+x" (res) : );
} }
template<> EIGEN_STRONG_INLINE Packet2d ei_pset1<double>(const double& from) { template<> EIGEN_STRONG_INLINE Packet2d ei_pset1<double>(const double& from) {
Packet2d res = _mm_set_sd(from); Packet2d res = _mm_set_sd(from);
return _mm_unpacklo_pd(res,res); return _mm_unpacklo_pd(res,res);
// asm("unpcklpd %[x], %[x]" : [x] "+x" (res) : );
} }
#else #else
template<> EIGEN_STRONG_INLINE Packet4f ei_pset1<float>(const float& from) { return _mm_set1_ps(from); } template<> EIGEN_STRONG_INLINE Packet4f ei_pset1<float>(const float& from) { return _mm_set1_ps(from); }